Technical field: The invention relates to a procedure, where among other measurements temperature sensors are used directly in the refining zone for linked refiners to minimize the risk to get operating conditions which can cause unacceptable pulp quality.

The procedure means that the pulp quality variations can be minimized as the pulp quality variables can be controlled by using an inner loop comprising two or more control objects in terms of a primary refiner and a secondary refiner or more refiners.

This differs from a traditional concept where each refiner has its own control concept.

The outer loop is designed to facilitate control of the pulp quality variables simultaneously. In those cases where the pulp quality is not measured properly a selective estimation of the pulp quality is obtained dynamically by using a model which uses the temperature measurements in the refining zones. The estimation provides an inter-sampling which makes it possible to design faster control concepts compared with traditional concepts.

The present invention is applicable in all technical areas where refiners are used, such as pulp and paper industry as well as related industries.

Technical background: Refiners of one sort or another play a central role in the production of high yield pulp and for pre-treatment of fibers in paper-making for the pulp and paper industry and related industries through grinding, for example, thermo-mechanical pulp (TMP) or chemical thermo-mechanical pulp (CTMP) starting from lignin-cellulose material such as wood chips. Two types of refiners are important to mention here; low consistency (LC) refining where the pulp is refined at about 4 per cent consistency (dry content), and high consistency (HC) refining where the consistency is commonly about 40 per cent. LC refining is done in a two-phase system chips/pulp and water, while HC refining has three phases; chips/pulp, water and steam. Refiners are also used in other industrial applications, such as for example manufacturing of wood fiber board. Most refiners consist of two circular plates, discs, in between which the material to be treated passes from the inner part to the periphery of the plates, see Figure 1. Usually, there is one static refiner plate and one rotating refiner plate, rotating at a very high speed.

The static refiner plate is placed on a stator holder (3), and is pushed towards a rotating one place on a rotor holder (4), electro mechanical or hydraulically (5).

The chips or fibers (6) are often fed into the refiners together with the dilution water via the center (7) of the refiner plates and are grinded on its way outward to the periphery (8). The refining zone (9), between the plates (also called segments) has a variable gap (10) along the radius ( 1 1) dependent on the design of the plates. The diameter of the refiner plates differ dependent on size (production capacity) of the refiner and brand. Originally the plates (also called segments 12, 13, see Figure 1 and Figure 2) were cast in one piece, but nowadays they usually consist of a number of modules (forming a disc) that are mounted together on the stator and rotor. The segments can be produced to cover the entire surface from the inner to the outer part of the holders or be divided into one inner part (14) often called "the breaker bar zone" and an outer part (15) called periphery zone.

These segments have grinding patterns (16), with bars (17) and troughs (18) that differ dependent on supplier. The bars can be seen as knives that defibrillate chips or further refine the already produced pulp. The plates wear continuously during the refining process and have to be replaced at intervals of around every 2 months or so. In an HC refiner, fibers, water and steam are also transported in the troughs between the bars. The amount of steam is spatially dependent, which is why both water and steam may exist together with chips/pulp in the refining zone. In an HC refiner water will normally be bound to the fibers. Dependant on the segment design different flow patterns will occur in the refiner. In an LC-refiner no steam is generated and thereby only two phases exist (liquid and pulp).

There are also other types of refiners such as double disc, where both plates rotate counter to each other, or conic refiners. Yet another type is called twin refiners, where there are four refiner plates. A centrally placed rotor has two refiner plates mounted one on either side, and then there are two static refiner plates that are pushed against each other using, for example, hydraulic cylinders thus creating two refining zones. When refining wood chips or previously refined pulp the refiner plates are typically pushed against each other to obtain a plate gap (10) of approximately 0.2-0.7 mm dependent on what type of refiner is used. The plate gap is an important control variable and an increased or reduced plate gap is performed by applying an electro mechanical or hydraulic pressure (5) on one or several segments dependent on the type of refiner. With that an axial force is applied on the segments. The force which acting in opposite direction to the axial force consists in HC-refining processes by the forces obtained from the steam generation and the fiber network. In those cases LC-refining is considered the axial force is neutralized by the forces extracted from the increased pressure in the water (liquid) phase and the fiber network. If the plate gap is changed for example 10%, the pulp quality is changed considerably. One system available on the market today is based on temperature measurements along the radius in the refining zone to visualize the temperature profile (19) alternatively the pressure profile (20) for control purposes, see Figure 3. For LC- refining the pressure is preferred to be measured but for HC-refining the temperature profile will be enough to measure.

When changing the process conditions in plate gap, production and the amount of added dilution water, the temperature is changed which gives an opportunity to control it in different ways. Several temperature- and/or the pressure sensors are often used and can be placed directly in the segments alternatively mounted in a sensor array holder (21) which can be placed between the segments (12 and 13), see Figure 1 , Figure 2 and figure 4 as described in EP 0788 407. Usually, the sensor array holder is implemented between two segments in the outer part, see Figure 2.

The design of the segments has proven to be of great importance for characteristics of the temperature profile along the radius. Therefore, it is difficult in advance to decide where the temperature sensors (22) and/or the pressure sensors (22) should be placed in the sensor array holder (21).

In the outlet from the refiners, preferable the primary refiner, a near infra red measurement unit is sometimes installed. This unit measures the pulp consistency and is used for controlling the flow of dilution water to the refiner.

The pulp quality is not measured in the blow-line from each refiner. Instead the pulp quality is normally measured after a large chest called the latency chest. This makes it possible to measure the pulp quality about 20-30 minutes after the treatment refiners.

In the literature, temperature measurements have shown to be an unusual robust technology for HC-refining control US2000/6024309. When changing the production, dilution water and the hydraulic pressure the temperature profile is changed dynamically. This dynamic change is visualized in Figure 5a, where a step change in dilution water affect the temperature profile in different ways dependent on where on the radius (11) we consider the process. It is seen that when the dilution water increase, the temperature (23) will decrease before the maximum (24) but increase (25) after the maximum. The reason is that the water added cools down the back- flowing steam at the same time as the steam which is going forward is warmed very fast.

When the production is increased the entire temperature profile (19) is lifted to another level (26), see Figure 5b. The same situation is valid when the plate gap (10) is reduced by increasing for instance the hydraulic pressure.

The non-linearities are affected also by the design of the segments. This can result in different temperature profiles (19, 32) and pressure profiles, see Figure 5c.

In traditional control concepts for refiner control, the specific energy, E, i.e. the ratio between the refiners motor load and the chip feed rate (30) F _P , alternatively only the motor load, the consistency, C for each refiner below indicated by using the sub- indices p for the primary refiner and s for the secondary refiner, see below. Pulp quality related variables (37), e.g. the Canadian standard freeness, CSF, which normally are analyzed after the latency chest (38), see Figure 6a which shows schematically a flow sheet, is normally controlled manually without automatic control concepts. The elements in the output vector Y are thereby affected by the elements in the input vector U which often comprises hydraulic pressure (5) Phydr, dilution water feed rate (29) F _D , and the chip feed rate (30) F _P dependent on the refiner to study, i.e. the primary or the secondary refiner.

where G represents the transfer function matrices with its elements g, describing the dynamics in the system. The pulp properties (31) out from each refiner are not controlled and varies dependent on the conditions inside the refining zones. The linear function G represents a simplification of the process dynamics as it is strongly non-linear which will be commented below.

The refining processes are often designed by using two serially linked refiners; one called primary stage (34) and one called secondary stage (35), and often also a process stage called a reject refiner (36), see Figure 6a. Sometimes another structure is used with parallel designs which make the control concepts more complex.

Examples of control concepts available on the market today is presented in the thesis from the Mid university in Sweden, "Quality Control of Single Stage Double Disc Chip Refining ", Joar Liden, 2003 and US2005/0263259 where the control concept is based on a Model Productive Controller, MPC, which is used for large complex systems comprising several refiner lines but also single refiner lines and refiners.

These concepts are not based on measurements obtained directly from the refining zones as described in US2000/6024309. Instead these concepts are focused on available process variables which are measured outside each refiner.

In some research projects prediction of pulp quality out from the refiner lines has been performed off-line. The method used has been based on "Auto Regressive Moving Average eXogenous"(ARMAX) modeling procedures, for details see "System identification, Theory for the user", Lennart Ljung, 2nd edition, Prentice Hall, New Jersey (1999), and can be seen as a subset of a number of system identification tools available on the market today. The off-line trials resulted in an article "Refining zone temperature control: A good choice for pulp quality control? ", Karin Eriksson och Anders Karlstrom, IMPC09, 2009 where the dynamic effects on the pulp quality were studied by using a new type of ARMAX-modeling procedure which still can be characterized as a state space model which is easy to translate to transfer functions G if required. The aim was to investigate if an empirical correlation exists between the pulp quality and step changes in dilution water flow rate, production and hydraulic pressure from a refiner. The primary result obtained was that the prediction of the pulp quality was slightly better when information from the temperature profile in the primary refiner was included in the vector U.

The above mentioned descriptions about what actually happens inside the refining zones at different process conditions has been included to introduce what we call direction dependent dynamics. Direction dependent dynamics means that the pulp quality from different step changes in the input signals (5, 29, 30, 31), see Figure 6b, will be different dependent on the direction of the step, negative or positive step, see "Realisation and estimation of piecewise-linear output-error models ", Rosenqvist and Karlstrom, Automatica (2005). In some process operating point the direction dependency not importnat while in other operating points it will be central.

The direction dependent dynamics reflect a non-linear dynamics which can cause problems when designing control systems.

Sometimes the process dynamics is described as a non-linear function f _m instead of the linear description g _m . Simplified the function ,, can be described as Jm ~

i.e. a non-linear function (with the time constants T _mt and T _m2 and the gains k _ml , k _m2 respectively) which approximately can be describes as two linear transfer function where t and i describes the direction of the steps in the elements u _m in the input vector U.

All refiners are different dependent on construction, type of segment and non- linearities in the process. Therefore, a simplified analysis in terms of step changes in both directions should be performed when introducing advanced control concepts. The reason is to confirm which element f _m to neglect and investigate the degree of direction dependency. In a research project a new theoretical physical model has been documented ("Refining models for control purposes" (2008), Anders Karlstrom, Karin Eriksson, David Sikter and Mattias Gustavsson, Nordic Pulp and Paper journal). The model, describes HC -refining and it is supposed that the temperature and/or the pressure are measured along the radius of a segment to span the material and energy balances in the refiners and thereby make it possible to estimate the plate gap. The main difference compared with earlier rudimentary trials to describe the physics of the grinding processes is that the model estimates both the reversible thermodynamic work and the irreversible defibration work applied on the fiber network where the shear forces have a central position when iterating to find the right plate gap. Thereby, the model is described from an entropy perspective instead of an enthalpy based approach which does not take into account the shear between the fibers, flocks, water and the segments. In this model it is assumed that the production rate is possible to measure which indeed is hard to prove due to for example the fluctuations in raw material density et cetera. The model shows that variations in the temperature profile also affect the local consistency estimation which results in a variable irreversible defibration work and hence correlates to the final pulp quality.

Disclosure of the invention Technical problem A comprehensive material has been presented in the literature regarding control of refiners using dry content measurement, plate gap measurement, and temperature measurement including safety systems for preventing plate clash of segments. Documents regarding control of refines using pulp quality measurement are however surprisingly underrepresented.

Measurement systems for pulp quality are often constituted by both hardware, such as sampling devices from the process, and image analysis systems for fiber characterization. The latter is connected to software that calculates pulp fractions which constitute for example mean fiber length (MFL) calculations. The Canadian Standard Freeness (CSF) is usually also given by the measurement systems for pulp quality. Some results indicate that measurement of pulp quality variables using such instruments show a clear deviation from actual conditions achieved during the direct refinement of the fibers. This may, for example, be caused by two serially connected refiners, often denoted primary and secondary refiner, may be dynamically operating in a number of different ways, while at the same time various operating points may be used and this has not been taken into account in previously published patents. Deviations in the pulp quality measurement may also be caused by local fluctuations in the latency tank (38) in Figure 6a which causes inhomogeneous pulp makes the control of the quality harder.

For a long time it has been thought that it is possible to control the character of the pulp, and thereby indirectly the final pulp quality after the secondary stage, using only the temperature profile and/or the pressure profile out of the primary refiner. This is obviously a simplified truth as the pulp quality is set by the conditions in both the primary and the secondary refiners. Research also shows that the temperature profile responds differently depending on the refiner, the grinding segment patterns and on where in the refining zones the fibers have been refined, that is related to the local retention time. This means that one has to connect the temperature profiles and/or the pressure profiles in a more elaborate way to the final pulp quality variables, as compared to previous patented suggestions.

Another problem that previously has not been considered is that, despite a smooth outgoing dry content in the blow-line pipe coming from the refiners, the local dry content along the radius of the refining zone may vary considerably. This obviously affects the final pulp quality as the retention time of the fibers thereby varies, which makes it insufficient to measure the dry content coming out from the primary refiner and/or secondary refiner.

It has also been believed that the characteristic pulp properties obtained from the primary and secondary refiners may be controlled using a well designed pulp quality sensor that gives absolute values. We have however found that many control systems that measures absolute variables, sometimes are not used for on-line control at all in TMP-process lines, which is a result of uncertainties in the measurement and of the sampling rate being slow with sometimes up to 25 minutes between measurements, causing proposed control concepts to not being used for control in practical cases. Figure 7 shows a typical example of measured CSF (39) as a function of time, and it is clearly seen that the variation as compared to the sliding average of CSF (40) is approximately +/- 20 ml (41). A more reliable CSF signal is usually needed than the case in Figure 7. Figure 8 shows a signal (42) averaged over five samples (corresponding to 125 minutes) with a variation of +/- 7 ml (43) which is more acceptable. As seen from the perspective of control, having such time horizons to handle is unacceptable when the pulp quality in the refining zones is set within a second. In addition to the pulp quality characterization being uncertain, there are also other problems when control related technical solutions are to be formulated. A central issue that has to be addressed is the long lead times in the process. The lag time for a typical system with two serially connected refiners and an adjoining latency chest may be anything from 20 to 30 minutes depending on the process design. This implies that a change in the refining conditions in the refining zone of one of the refiners has no effect on the measured pulp quality until in 20-30 minutes. This may potentially be solved with a more elaborated process design, but one usually wants a latency chest between the refiner line and the pulp quality analyzer. How to handle the inhomogeneous pulp, caused by an uneven control of the refining properties, is thus central to the disclosure of the technical problem. If the packing degree of the pulp in the refining zone varies locally over time and space, this may create local areas where the spatial temperature or pressure may facilitate a solution. All process conditions, such as increased production or dilution water flow rate, that affects the active volume in the refining zone at constant hydraulic pressure, therefore affect both the plate gap and the shape of the temperature and/or pressure profile according to the previous description, see Figure 5. This makes the pulp retention time in the refining zone vary, which affects the fluctuations in the refining zone and finally the pulp quality at normal operation. The process properties may also be negatively affected such that the refiner drifts away against operating points that are prohibited due to risk of malfunction. The prohibited areas are hard to predict in advance using contemporary technology, which causes these problems not to be detected until in about 25 to 125 minutes, depending on the sampling rate and reliability of the pulp analyzer. A partial solution to the problem has been to control the temperature profile in the refining zones, as has been pointed out in a number of previous patents, but it has so far been unknown how the temperature profiles are related to the achieved pulp quality. In conjunction with previously mentioned ARMAX modeling attempts on a primary refiner, it turned out to be difficult to know which input signals are necessary for achieving a dynamic follow up of the pulp quality as the process line comprises several refiners, but it became clear that the temperature profile information gives additional value as input signals when predicting the pulp quality. Another problem that occurred during the system identification attempts is that the models only describe the dynamical response to changes in the process, and thus knowledge about the absolute values of the pulp quality levels is lacking. We therefore get a bias problem, that is the estimated signals deviate from the absolute value of the measured output signal, while we concurrently have to handle two time scales, one for the fast dynamics following form the fast profile measurements in the refining zone, and one for the slow dynamics caused by the slow sampling rate of the pulp quality analyzer. The ARMAX models are, as previously mentioned, linear state models and this gives problems when non linear processes with directional dependent dynamics is to be identified. It has also turned out to be difficult to handle slower variations, trends, in the process that most often are dependent on non linear phenomena such as wear in grinding segments etc.

Yet another problem that has not been penetrated in the literature is how the linear systems identification methods, for example the ARMAX models, may recreate reliable time constants and amplifications that may be used for control purposes when directionally dependent dynamics is present. How the identification of the pulp quality variables optimally should be performed, and how this recently acquired knowledge is to be used for inter-sampling pulp quality via soft sensors, that is algorithms

(empirical models) that makes it possible to speed up the sampling rate, from today's 25 minutes to approximately 1-2 minutes, has hitherto neither been known, and this has resulted in the technical problem not being solved previously.

No results have yet been published where both the primary and the secondary refiners are provided with sensor arrays for temperature and/or pressure measurement, which has proven to constitute an important part of the technical solution below. The solution

The present invention constitute a solution to these problems and relates to a method that uses robust temperature and/or pressure measurement directly in the refining zone combined with accessible signals from the process and a model for estimating the spatial dry content in each refining zone and/or the estimated pulp quality in order to control a complete refiner line.

As the measurement sensors are positioned along the radius in the refining zone, a temperature vector is formed that constitutes the so called temperature profile, see Figure 5. In the case pressure sensors are used it is then called pressure profile.

In order to get a good characterization of the pulp quality it is not enough to measure the refining zone temperature in one point or the temperature profile in the primary refiner only. Instead, the temperature and/or pressure profiles from both the primary and secondary refiners must be used as both refiners affect the pulp quality. In certain cases the reject refiner may be included if it significantly affects the end product, that is the pulp quality. PHYSICAL MODEL FOR SPATIAL DRY CONTENT MEASUREMENT

In case a dry content measurement is made in the blow-line from the refiners, it is preferably controlled by the dilution water flow rate. The physical model presented in "Refining models for control purposes " (2008), Anders Karlstrom, Karin Eriksson, David Sikter and Mattias Gustavsson, Nordic Pulp and Paper journal has turned out to be useful for calculating the dry content output from the refiners. A comparison between measured and predicted dry content shows that the predicted dry content correspond better with laboratory samples taken from the blowout pipe.

The measuring device in the blow-line is unable to handle variations in dry content locally in the refining zone. Instead, the physical model must be used to access these internal states along the radius of the refining zone. This therefore means that the physical model gives access to the dry content profile in the refining zone. The model does though assume that the energy balance in the process uses the temperature profile or pressure profile as input variable in the calculation. A large change in , for example, the inner part of the temperature profile gives a large change in the estimated dry content in the same region, see Figure 5. EMPIRICAL MODEL

Through the access to temperature profiles and/or pressure profiles from the primary and secondary refiners, in conjunction with traditional process variables such as chip flow rate, dilution water flow rate, hydraulic pressure and inlet pressure, that in combination may constitute the input signal vector u(t), we are able to study how the pulp quality in the vector y(t) relates to various changes in the process, assuming a reliable model can be created.

The reason why this is of interest is obviously that we would like to couple, via empirical models, process variables and refining zone information to several pulp quality variables, such as for example CSF and MFL, that may be used for control purposes.

We will not go too deep into details regarding all types of empirical models that may be used, but in case we use a linear model, it is important to show an example so what we want to achieve is comprehended.

It is important to point out that the model

A(q)y(t)=B(q)u(t-nk)+C(q)e(t), is an "Auto Regressive Moving Average eXogenous"(ARMAX) -based equation where the output signal vector y(t) contains the signals we want to model. The input signal vector u(t) can, as mentioned above, for example contain some variables from the temperature profile of the refining zone, production, dilution water flow rate, and hydraulic pressure, that also may be replaced by the plate gap if it is measured, nk represents lag time for respective input signal to the output signal, e(t) is the error and A(q), B(q) and C(q) are polynomials in the operator q ^l .

It should be noticed than the model only handles the variations, which means that it is not the absolute value that is modeled. The output signal vector y(t) must therefore be coupled to the slowly varying part (40) of the signal that is being modeled in order to achieve the absolute value of the signal that is being modeled, such as the pulp quality variables CSF (39) or other dynamic characteristic measurement signals, see Figure 7. Often the slowly varying part of the signal is constituted by an average or a heavily filtered linear trend, see "System identification, Theory for the user ", Lennart Ljung, 2nd edition, Prentice Hall, New Jersey (1999).

The input signal vector u(t) has a tendency to become complex and vast for cases where complete refiner lines are viewed, so it is preferably divided into vectors for the process variables V _(prim) (44) and V _(sec) (46), while measurement signals from the refining zones are represented by the vectors T _(prim ) (45) and T _(sec) (47), see Figure 9a. Another reason for this subdivision is also that sometimes not all information is needed as input signals in making the model. For certain types of refiners it is enough to only have measurement signal from the refining zone T _(prim ) (45) och T _(sec) (47). In other cases it may be important to also include the process variables V _(prim) (44) and V(sec) (46), especially if the operating points of the refiner change over time intensely.

In order to generate the dynamical empirical models in Figure 9a, ARMAX- models or other models may for example be used. The output signals (48, 49) from each model in combination with the slow part of the signal (51) may be added and constitute the final pulp quality vector (50). Note that good models are achieved only if the retention time in the latency chest is taken into consideration, as a sufficient excitation of an input signal in the refiners will show in the measured pulp quality no earlier than in 20 minutes. In the identification of the output signal y(t), the lag time nk must be included in order to get a good dynamic adherence to the measured pulp quality. MODEL IMPROVEMENTS OF y(t)

In the model that describes the pulp quality output from the refiner, we do however want to find estimates significantly faster than what is achieved with the previous model in Figure 9a. In Figure 9b an example is shown. The actual variation in pulp quality output from the latency tank (52) is modeled and achieved as the sum of the variation in pulp quality from the primary (48) and secondary (48) refiners, and is illustrated by (53). If we exclude the lag time nk from the model, the pulp quality (54) is achieved. If we further shorten the settling time processes caused by stirring phenomena and natural filtration in the latency tank a modified output signal y _mO d( (55) is achieved, that represents the pulp quality variation significantly faster than the pulp quality analyzer may achieve.

MODEL IMPROVEMENTS TO THE SLOWLY VARYING SIGNAL

The modified output signal y _mo d(t) should then be coupled with a modified slowly varying signal in order to get a new estimated pulp quality without lag time, see Figure 10. When there are small variations in the process, it is normally small variations in the pulp quality, but not in the measured pulp quality. Therefore, a slowly varying signal (56) is used, created for example from the averaged signal (42) in Figure 8 or a signal averaged over 2-10 samples, depending on which filtration is preferred.

This, by using the averaged signal 2-10 samples backwards in time as the slowly varying part (56) in combination with the modified output signal y _mo d(t) (57) that has a fast variation, a more reliable value is achieved where two time scales are connected to each other and makes use of the speed, while we can follow the long term trend in the measurement signal from the pulp quality analyzer, see Figure 10.

For non linear systems, specific tests may be performed in order to create models that handles so called direction dependent dynamics, see Figure 6b, where the function^ describes the response of the system to various changes in the input signals u(t).

Which system architecture that is used to predict the pulp quality depends on the accuracy needed to control the process. In many cases a linear model may work sufficiently well, while other situations need non linear solutions. It should be noted in this context that the previous identification technique for finding a god state model, which may then be transformed into the system of transfer functions below, may be used to estimate the gains and the time constants according to the equation

by restructuring (segmenting) the measurement series in order to capture dynamics that describes† and i .

It is thus important to note that the measurement of temperature, alternatively pressure or a combination of the two, along the radius of the grinding segments is necessary in order to be able to estimate the pulp quality during HC refining or LC-refining.

Due to the access to rapid information from for example temperature sensor in the refining zones of the primary and secondary refiners, in combination with a sufficiently good model, it is thus possible to get a fast prediction of the pulp quality variables. Therefore, the quality of the pulp can be prevented from drifting outside the specifications due to changes in the process conditions. This is preferably done by inter-sampling the pulp quality (57) for example every minute, see Figure 10, so that it is possible to use the predicted pulp quality in various control concepts. At the same it is important that the predicted pulp quality (57) is continually compared with the measured pulp quality (39) even if the estimated quality usually is more precise and better suited for control purposes.

Figur 1 1 a schematically shows the control of general processes with a traditional control concept, where a controller (C) is provided with the difference between a set point value (SP) and a measured signal (M) and then via a device which controls a process (//). Figure l ib schematically shows the control of the process using the new control concept. The controller ( ), constituted by a computer or similar electronic equipment, is fed with the difference between (the set point values (SP)) and (the actual values (PV)) of the fast estimated pulp quality {M _moc j). The controller ( ) gives information to a distribution block (D) which divides the control signal to the interior control loops for each refiner that constitute part of the control concept. The interior control loops may be constituted by a number of interior control loops for controlling the spatially estimated dry content using the dilution water flow rate and/or the measured temperature profiles and/or the pressure profiles in the primary and secondary refiners, respectively. The temperature profiles and/or pressure profiles preferably control the hydraulic pressures (5) in respective refiner, but also in combination with supplied water (43) or chip supply (6). Relevant variables from the measurement sensors is the plate gap in combination with a modeled dry content at various positions in the plate gap are then fed to a model block that sums up the final estimated variation of the pulp quality (58). This signal is added to the measured signal M via a filter Mf _llt if the signal qualitative of M varies a lot over time.

A number if distribution routines may be formulated and one is given in Figure 11c, where two serially connected blocks K and R describes a distribution set by an operator and R is a routine that relates to the actual gain distribution for the refiner zones measurements/estimations at normal production. Using K, the operator may thus increase or diminish the effect of R on respective internal control loop. The internal loop assumes measurement of some physical quantity in the refining zone, see Figure l id. Each control loop has its own controller C to handle the dynamics H that de facto may vary from refiner to refiner. The model, see Figure l ie may be described in various ways, but a common feature of all models is that is deals with fast variations on the same time scale as the internal control loops. The output signals from each partial model F (59) are summed up and forms (55) that is then summed up with a trend signal (60), see Figure l ib and Figure l lf. The final modifying signal {M _mod ) is then fed back to create the necessary control difference that enters the controller ( ). Note that in Figure l lf an arbitrary number of serially connected refiners and/or refiners connected in parallel that are controlled by the valid control concept.

In case a sufficiently accurate plate gap measurement gauge is available, the calculation method above may include that too as a follow up.

The main purpose of the invention is thus to disclose a method that with a large degree of accuracy is able to present an on-line based estimation of pulp quality while it concurrently is used for controlling the pulp quality output from the last refiner and then the latency chest. The invention is based on that the temperature profile and/or the absolute pressure profile can be measured in the primary and/or the secondary refiner's refining zones and/or that the spatial dry content in the refining zone can be estimated using a model.

The method according to the present invention is not restricted to any specific device for sensing temperature or pressure in the refining zone. Such devices are however known through for example the Swedish patent 9601420-4. The invention of neither limited to the illustrated embodiment, but may be varied in various ways within the scope of the claims.

Deskription of the drawings:

Figure 1 : Cross section of a stationary disc which is pushed towards a rotating disc. Figure 2: Two segments where the sensor array holder, used for temperature- and/or pressure measurements, is placed in between.

Figure 3 : Temperature profile and pressure profile as a function of the radius in the refining zone.

Figure 4: Array shaped as a parallelepiped with discrete temperature and/or pressure sensors.

Figure 5a: The shape of the temperature profile before and after an increase in the dilution water feed rate.

Figure 5b: The shape of the temperature profile before and after an increase in production. Figure 5c: The shape of the temperature profile before and after a change in segments.

Figure 6a: Simplified view of two refiners connected to a latency chest with

subsequent pulp quality analyzer. A reject refiner is also included in the figure. Figure 6b: Example of directional dependent dynamics.

Figure 7: Example of a measured pulp quality variable in the form of CSF.

Figure 8: Example of a signal averaged over 5 samples based on the signal in Figure 7.

Figure 9a: Simplified view of how the modeling blocks are connected and summed to represent the variation in the pulp quality. Figure 9b: Simplified figure for describing how a modified pulp quality variables behavior changes as lag time and time constants are modified to better represent what happens in each refiner.

Figure 10: Example of an averaged signal (DC-level) and a signal with a fast variation (AC-level) which are to be added and represent the final pulp quality. Figure 11a: Traditional control loop concept.

Figure l ib: New control concept according to this patent.

Figure 11c: Distribution block distributing the output signal from the regulator to respective internal control loop.

Figure l id: Example of an internal control loop with an arbitrary number of refiners that are connected in series or in parallel.

Figure l ie: Example of a model block with subsequent summation for final pulp quality variation. Figure 1 If: Compilation of the entire control concept in its detail.